Table 1 Overview of key distinctions in pre-processing strategies between domain non-specific images (e.g., natural images) and domain-specific images (e.g., medical images).

From: The impact of pre-processing techniques on deep learning breast image segmentation

Parameters

Domain non-specific

Domain specific

Orientation

No standardization

Same orientation

Pixel Spacing

No standardization

Median value of each axes

Pixel Intensity

Minimum–maximum

Percentiles, Z-Score,

Histogram Normalization,

and others

Image Size

Resize to square shapes

Resize maintaining aspect ratio

OR

Padding and resizing

  1. The table highlights differences in handling image size, pixel intensity, pixel spacing, and orientation.